3.1 Operating System

Basically, about 50% of the developers use Windows, and the users using Linux an MaxOS are almost the same. The group using BSD/Unix is relatively small.

os <- survey %>% 
  select(Country, DevType, Age, OperatingSystem, Gender) %>% 
  mutate(DevType2 = split_fn(DevType),
         gener2 = split_fn(Gender))  %>% 
  group_by(OperatingSystem) %>% 
    mutate(os_n = n()) %>% 
  ungroup() %>% 
  filter(!is.na(OperatingSystem))

os1 <- os %>% 
  select(OperatingSystem, os_n) %>% 
  distinct() %>% 
  mutate(value = os_n / dim(os)[1] * 100) 

os1 %>% 
  ggplot(aes(x = "", y = value, fill = factor(OperatingSystem))) +
  geom_bar(width = 1, stat = "identity") +
  coord_polar(theta = "y", start = pi/3) +
  scale_fill_manual(values = colors[c(2,1,5,6)])  +
  geom_text(aes(y = c(90, 25, 65, 0), 
                label = percent(value/100)), size = 5, color = "white") +
  blank_theme +
  theme(axis.text.x = element_blank()) +
  labs(fill = "OS")

4.1 How many people are employed full-time?

employ <- salary %>% select(Employment, emp_n) %>% 
  distinct() 

employ %>% 
  mutate(count = emp_n / sum(emp_n) ) %>% 
  ggplot(aes(x = reorder(Employment, count), y = count)) +
  geom_bar(stat = "identity", fill = colors[5], width = 0.5) +
  geom_label(aes(x = Employment, y = 0.1, 
                label = str_c("(", round(count*100, 1), "%)")),
            fontface = "bold", hjust = 0.5, vjust = 0.5, 
            size = 4, color = "black", fill = colors[3]) +
  scale_y_continuous(labels = percent_format()) +
  coord_flip() +
  labs(x = "", y = "Number of Participants",
       title = "Employment Status") +
  theme(axis.text.x = element_text(size = rel(1.25)),
        axis.text.y = element_text(size = rel(1.25)),
        plot.title = element_text(size = rel(1.25), face = "bold"))

Among all the pariticipants, 71.3% of them are full-time employed.

4.2 Who did not report salaries?

## The maximum salary is from a back-end developer
## The wealthest person currently is Jeff Bezons, who has 134.3 billion US dollors. 134.3 billion USD = 1.343e+11
## If the salary in the data is higher than this number, it was filtered. But the first thing to do is to unify the currencies, let's do it in US dollars.

fn_student <- function(x) {
  split_fn <- function(y) str_split(y, pattern = ",")[[1]][[1]]
  plyr::ldply(x, split_fn)[[1]]
}

no_salary <- salary %>% 
  filter(ConvertedSalary < 1) %>%
  mutate(Student2 = fn_student(Student),
    stdt_grp = if_else(Student2 == "No", "N", "Y")) %>% 
  group_by(stdt_grp) %>% 
  mutate(student_n = n()) %>% 
  ungroup() %>% 
  group_by(Country) %>% 
  mutate(country_n2 = n()) %>% 
  ungroup()

In fact, only about 49% of the participants who reported zero salary are student.

## To get a more compact view, the countries with 
## less than five participants were filtered out.
no_salary %>% 
  select(Country, country_n2) %>% 
  filter(country_n2 > 5) %>% 
  distinct() %>% 
  mutate(count = country_n2 / dim(survey)[1]) %>% 
  ggplot(aes(x = reorder(Country, country_n2), country_n2)) +
  geom_bar(stat = "identity", fill = colors[5]) + 
   # geom_label(aes(x = Country, y = 1e-4, 
   #              label = str_c("(", round(count*100, 2), "%)")),
   #          fontface = "bold", hjust = 0.5, vjust = 0.5, 
   #          size = 4, color = "black") +
   # scale_y_continuous(labels = percent_format()) +
  coord_flip() +
  labs(y = "", x = "", title = "Who Did not Provide the Salary?") +
  theme(axis.text.y = element_text(size = rel(1.15), angle = 15),
        axis.text.x = element_text(size = rel(1.15), angle = 15),
        plot.title = element_text(size = rel(1.25), face = "bold"))

4.3 Salary distribution by developer types

employed <- "Employed full-time"
salary1 <- salary %>% 
  filter(Employment == employed, ConvertedSalary > 1000) %>%
  mutate(Student2 = fn_student(Student),
    stdt_grp = if_else(Student2 == "No", "N", "Y")) %>% 
  group_by(stdt_grp) %>% 
  mutate(student_n = n()) %>% 
  ungroup() %>% 
  group_by(Country) %>% 
  mutate(country_n = n()) %>% 
  ungroup() 

salary_median <- median(salary1$ConvertedSalary)

options(scipen = 10000)
salary1 %>%
  filter(Gender %in% c("Male", "Female")) %>% 
  ggplot(aes(x = DevType2, y = ConvertedSalary, fill = Gender)) +
  geom_boxplot(fill = colors[5], outlier.shape = 21, outlier.size = 0.5) +
  scale_y_log10(labels = scales::comma,
                breaks = c(1e4, 1e5/2, 1e5, 1e6)) +
  coord_flip() +
  labs(y = "Annual Salary (USD)", x = "",
       title = "How Much Were Different Types of Developers Paid?") +
  geom_hline(yintercept = salary_median,
             color = colors[1], linetype = "dashed", size = 1) +
  scale_x_discrete(limits = salary1$Country) +
    theme(axis.text.x = element_text(size = rel(1.15), angle = 30),
        axis.text.y = element_text(size = rel(1.15)),
        axis.title.x = element_text(size = rel(1.15), face = "bold"),
        plot.title = element_text(size = rel(1.25), face = "bold"))

vline <- function(x = salary_median, color = colors[1]) {
  list(
    type = "line", 
    y0 = 0, 
    y1 = 1, 
    yref = "paper",
    x0 = x, 
    x1 = x, 
    line = list(color = color)
  )
}

salary1 %>% plot_ly(y = ~DevType2, x = ~ConvertedSalary, 
                    line = list(color = colors[5]),
                    marker = list(size = 3),
                  type = "box") %>% 
  layout(title = "Salary Distribution by Developer Type",
         xaxis = list(title = "Annual Salary (USD)", type = "log"),
         yaxis = list(title = ""),
         shapes = list(vline(salary_median)))

There are a total of 36014 participants answered this question. The median salary is 60391 USD. The median salary of Engineering managers, DevOPs specialists, C-suite executives are the highest.

4.4 How much does data scientists earn?

The figure is a salary distribution by countries and gender. Only countires with more than 200 participants were included. Similarly, the orange veritical line is the overall median salary (60391 USD).

salary2 <- salary1 %>% 
  mutate(gender2 = split_fn(Gender)) %>% 
  group_by(gender2) %>% 
  mutate(gender_n = n())

gender_type <- unique(salary2$gender2)


data_dev <- unique(salary2$DevType2)[c(2, 12)]

sal2 <- salary2 %>% 
  filter(DevType2 %in% data_dev, 
         gender2 != gender_type[3],
         country_n > 200, 
         Employment == employed)

sal2 %>% plot_ly(y = ~Country, x = ~ConvertedSalary, 
                    color = ~gender2,
                    line = list(color = colors[c(1, 2, 5)]),
                    marker = list(size = 3),
                 type = "box") %>% 
  layout(boxmode = "group",
         title = "Salary Distribution by Developer Type",
         xaxis = list(title = "Annual Salary (USD)", type = "log"),
         yaxis = list(title = ""),
         shapes = list(vline(salary_median)))

# sal2 %>% 
#   ggplot(aes(x = Country, y = ConvertedSalary, fill = gender2)) +
#   geom_boxplot(outlier.size = 0.75, outlier.shape = 21, width = 0.75) +
#   geom_hline(yintercept = salary_median, color = colors[6], 
#              size = 1.5, linetype = "dashed") +
#   scale_y_log10(breaks = c(1e4/2, 1e4, 1e5, 1e6/2, 1e6)) +
#   scale_fill_manual(values = colors[c(1, 3, 5)]) +
#   coord_flip() +
#   labs(x = "", y = "Annual Salary (USD)", 
#        fill = "", title="Annual Salary of Data Scientists/Analysts") +
#   theme(axis.text.x = element_text(size=rel(1.15), angle = 30),
#         axis.text.y = element_text(size = rel(1.15)),
#         plot.title = element_text(size = rel(1.25), face="bold"))

4.5 Data sicence salary distribution by gender

Only countries with more than 200 participants were included.

kable(matrix(gender_type, ncol = 1)) %>% kable_styling()

salary2 %>% 
  filter(country_n > 200, 
         DevType2 %in% dev_type[c(3, 8)],
         Employment == employed,
         ConvertedSalary > 1000,
         gender2 %in% c("Male", "Female")) %>%
  plot_ly(y = ~Country, x = ~ConvertedSalary, 
                    color = ~gender2,
                    line = list(color = colors[c(1, 2, 5)]),
                    marker = list(size = 3),
                 type = "box") %>% 
  layout(boxmode = "group",
         title = "Data Scientists/Analysts Salary Distribution by Countries",
         xaxis = list(title = "Annual Salary (USD)", type = "log"),
         yaxis = list(title = ""),
         shapes = list(vline(salary_median)))

  # ggplot(aes(x = gender2, y = ConvertedSalary, fill = gender2)) +
  # geom_boxplot(outlier.size = 1, outlier.shape = 21, width = 0.35) +
  #   geom_hline(yintercept = salary_median, 
  #            linetype="dashed", color=colors[6], size = 1.5) +
  # scale_y_log10(breaks = c(1e4/2, 1e4, 1e5/2, 1e5, 1e6)) +
  # scale_fill_manual(values = colors[c(1,3,5)]) +
  # # scale_color_manual(values = colors[c(1,3,5)]) +
  # # coord_flip() +
  # labs(y = "Annual Salary (USD)", x = "", 
  #      fill = "", title = "Data Scientist Annual Salary Distribution by Gender") +
  # theme(axis.text.y = element_text(size = rel(1.15)),
  #       axis.text.x = element_text(size=rel(1.15), angle = 25),
  #       axis.title = element_text(size = rel(1.15), face = "bold"),
  #       plot.title = element_text(size = rel(1.25), face = "bold")
  #       )

Overall, female developers seem to slightly better paied than male developers

4.6 Salary distribution by countries and gender

• By Country Only

sal4 <- salary2 %>% 
  filter(country_n > 100, Employment == employed)

sal4 %>% plot_ly(y = ~Country, x = ~ConvertedSalary, 
                    # color = ~gender2,
                    line = list(color = colors[c(6)]),
                    marker = list(size = 3),
                 type = "box") %>% 
  layout(boxmode = "group",
         title = "Data Scientists/Analysts Salary Distribution by Countries",
         xaxis = list(title = "Annual Salary (USD)", type = "log"),
         yaxis = list(title = ""),
         shapes = list(vline(salary_median)))

  # ggplot(aes(x = Country, y = ConvertedSalary)) +
  # geom_boxplot(fill = colors[5], 
  #              outlier.shape = 21, outlier.size = 0.5 ) +
  # scale_y_log10(breaks = c(1e4, 1e5/2, 1e5, 1e6/2, 1e6)) + 
  # coord_flip() +
  # geom_hline(yintercept = salary_median, 
  #            color = colors[1], linetype = "dashed", size = 1.5) +
  #   labs(y = "Annual Salary (USD)", x = "", 
  #        title = "Salary Distribution by Countires") +
  # theme(axis.text.y = element_text(angle = 10, size = rel(1.15)),
  #       axis.text.x = element_text(size = rel(1.15), angle = 30),
  #       axis.title = element_text(face = "bold", size = rel(1.15)),
  #       plot.title = element_text(face = "bold", size = rel(1.15))) 

# ggsave("~/Downloads/Images/salary-country.png", dpi=300,
#         width = 12, height = 10, unit = "in")

It appears that most of the millionare developers are from the US.

• Included Gender

salary2 %>% 
  filter(country_n > 100, gender2 == gender_type[c(1, 2)]) %>% 
  group_by(Country) %>% 
  mutate(med_salary = median(ConvertedSalary)) %>% 
  ungroup() %>% 
  arrange(desc(med_salary)) %>% 
  plot_ly(y = ~Country, x = ~ConvertedSalary, 
                    color = ~gender2,
                    line = list(color = colors[c(1, 6)]),
                    marker = list(size = 3),
                 type = "box") %>% 
  layout(boxmode = "group",
         title = "Data Scientists/Analysts Salary Distribution by Countries",
         xaxis = list(title = "Annual Salary (USD)", type = "log"),
         yaxis = list(title = ""),
         shapes = list(vline(salary_median)))

  # ggplot(aes(x = Country, y = ConvertedSalary, fill = gender2)) +
  # geom_boxplot(outlier.shape = 21, outlier.size = 0.5 ) +
  # scale_y_log10(breaks = c(1e4, 1e5/2, 1e5, 1e6/2, 1e6)) + 
  # scale_fill_manual(values = colors[c(1, 2, 5)]) +
  # coord_flip() +
  # geom_hline(yintercept = salary_median, 
  #            color = colors[6], linetype = "dashed", size = 1.5) +
  #   labs(x = "", y = "", fill = "", 
  #      title = "Salary Distribution by Countries and Gender") +
  # theme(axis.text.y = element_text(angle = 10, size = rel(1.15)),
  #       axis.text.x = element_text(size = rel(1.15), angle = 30),
  #       plot.title = element_text(size = rel(1.15), face = "bold"))

# ggsave("~/Downloads/Images/salary-country-gender.png", dpi=300,
#         width = 12, height = 10, unit = "in")

4.7 Does salary increase with length of experience?

# get_coding_year <- function(x) {
#    split_fn <- function(y) str_split(y, pattern = " ")[[1]][[1]]
# plyr::ldply(x, split_fn)[[1]] 
# }
# 
# year_up <- function(x) {
#    split_fn <- function(y) str_split(y, pattern = "-")[[1]][[2]]
# plyr::ldply(x, split_fn)[[1]] %>% as.numeric()
# }
# 
# year_low <- function(x) {
#    split_fn <- function(y) str_split(y, pattern = "-")[[1]][[1]]
# plyr::ldply(x, split_fn)[[1]]  %>% as.numeric()
# }

salary3 <- survey %>% 
  select(Country, ConvertedSalary, Age, Gender, Employment,
         YearsCoding, YearsCodingProf, DevType2) %>% 
  drop_na() %>% 
  mutate(gender2 = split_fn(Gender)) %>% 
  group_by(Country) %>% 
  mutate(country_n = n()) %>% 
  ungroup() %>% 
  as_tibble() 

age_sal <- salary3 %>% filter(Employment == employed, 
                   country_n > 200, ConvertedSalary > 1000)

## reorder age groups by theirs age
age_factor <- factor(age_sal$Age)
age_levels <- levels(age_factor)
age_factor <- factor(age_factor, levels = c(age_levels[7], age_levels[1], 
                                             age_levels[2], age_levels[3],
                                             age_levels[4], age_levels[5],
                                             age_levels[6]))
age_sal$age_factor <- age_factor
age_sal %>% 
  ggplot(aes(x = age_factor, y = ConvertedSalary)) +
  geom_boxplot(width = 0.25, fill = colors[5],
               outlier.size = 0.5, outlier.shape = 21) +
  scale_y_log10(breaks = c(1e4, 1e5/2, 1e5, 1e6/2, 1e6)) +
  coord_flip() +
  labs(x = "Age", y = "Annual Salary (USD)", 
       title = "Annual Salary Distribution by Age") +
  theme(axis.text.x = element_text(size = rel(1.15), angle = 15),
        axis.text.y = element_text(size = rel(1.15)),
        axis.title = element_text(size = rel(1.15), face = "bold"),
        plot.title = element_text(size = rel(1.15), face = "bold"))

Clearly, there is a trend of salary increase with developers’ ages until 65 years and older.

df_lang <- survey %>% 
  select(Country, ConvertedSalary, Age, 
         LanguageWorkedWith, FrameworkWorkedWith, DevType2, Employment) %>% 
  as_tibble() %>% 
  filter(!is.na(LanguageWorkedWith))

langs <- str_split(df_lang$LanguageWorkedWith, ";") %>% 
  unlist()  %>% 
  unique() 

5.1 Salary distribution by languages

median_sal <- median(lang_sal2$ConvertedSalary)

lang_sal2 %>% 
  group_by(languages) %>% 
  mutate(median_salary = median(ConvertedSalary)) %>% 
  ungroup() %>% 
  ggplot(aes(x = reorder(languages, median_salary), 
                         y = ConvertedSalary)) +
  geom_boxplot(fill = colors[5], outlier.shape = 21, 
               outlier.size = 0.25, width = 0.5) +
  scale_y_log10(breaks = c(1e4, 1e5/2, 1e5, 1e6/2, 1e6)) +
  coord_flip() +
  geom_hline(yintercept = median_sal, color = colors[1], 
             size = 1, linetype = "dashed") +
  labs(x = "", y = "Annual Salary (USD)",
       title = "Salary Distribution by Languages Used (USD)") +
  theme(axis.text.x = element_text(size = rel(1.15), angle = 15),
        axis.title.x = element_text(size = rel(1.15)),
        axis.text.y = element_text(size = rel(1.15)),
        plot.title = element_text(size = rel(1.15), face = "bold"))

The overall median salary is $61,000. It should be noted that since a developer could have used more than one type of languages, while only the salary is reported as a whole contribution of all languages used. Thus, the salary may be (for sure) mutiplely counted. A ideal case is that a developer has reported the portion of contribution to the salary of a language, but this is impossible.

lang_median_salary <- lang_sal2 %>% 
  group_by(languages) %>% 
  summarise(median_salary = median(ConvertedSalary),
            max_salary = max(ConvertedSalary),
            min_salary = min(ConvertedSalary)) %>% 
  arrange(by = median_salary) %>% 
  ungroup()

kable(lang_median_salary) %>% kable_styling()

lang_median_salary %>% ggplot(aes(x = reorder(languages, median_salary), 
                                  y = median_salary)) +
  geom_bar(stat = "identity", fill = colors[5]) +
  geom_hline(yintercept = median_sal, size = 1.5, 
             color = colors[6], linetype = "dashed") +
  geom_label(aes(y = 3e4, x = languages, label = str_c("(", median_salary, " USD)")),
             hjust = 0, vjust = 0.5, size = 4, fontface = "bold", 
             color = "black", fill = colors[3]) +
    coord_flip() +
    labs(x = "", y = "Annual Salary (USD)", 
         title = "Median Salary by Languages Used (USD)") + 
    theme(axis.text.x = element_text(size = rel(1.15)),
          axis.text.y = element_text(size = rel(1.15)),
          plot.title = element_text(size = rel(1.25), face = "bold")) +
    annotate("segment", y = 6e4+2000, yend = 7e4, x=1, xend=4, 
             colour=colors[2], size = 1, arrow = arrow(length = unit(0.15, "inches"))) +
    annotate("label", y = 7e4+7000, x = 5.5, 
             label = "Overall Median Salary\n($61K)",
             color = colors[6], size = 4, fontface = "bold") 

5.2 Languages Used by Data Scientists/Analysts?

ds_dev <- unique(lang_sal$DevType)[c(3, 8)]

ds_lang <- lang_sal %>% select(JavaScript:Ocaml, DevType) %>% 
  filter(DevType %in% ds_dev)

ds_lang2 <- ds_lang %>% 
  gather(JavaScript:Ocaml, key = "language", value = "IS_LANG") %>% 
  filter(IS_LANG == 1) %>% 
  group_by(language) %>% 
  mutate(n_lang = n()) %>% 
  ungroup() %>% 
  select(language, n_lang) %>% 
  distinct() %>% 
  mutate(count =  n_lang / sum(n_lang))
  
ds_lang2 %>% 
  ggplot(aes(x = reorder(language, count), y = count)) +
  geom_bar(stat = "identity", fill = colors[5]) +
  geom_label(aes(x = language, y = 1e-2, 
               label = str_c("(", round(count*100, 2), "%)")),
           fontface = "bold", hjust = 0.5, vjust = 0.5, 
           size = 4, color = "black", fill = colors[3]) +
  scale_y_continuous(labels = percent_format()) +
  coord_flip() +
  labs(x = "", y = "",
       title = " Languages Used by Data Scientists/Analysts") +
  theme(axis.text = element_text(size = rel(1.15)),
        plot.title = element_text(size = rel(1.15), face = "bold"))

Undebatablely, Python has now been the dominating language used in the field data science. Interestingly, C is ranked number 2 in the DS’s language list. Also, SQL is a must have tool for a data scientist/analyst.

5.3 Languages Desire Next Year

6.1 IDEs included in they survey are:

df_ide <- survey %>% 
  select(Country, DevType2, IDE, ConvertedSalary, Gender, Age, 
         YearsCoding, YearsCodingProf, LanguageWorkedWith, 
         FrameworkWorkedWith) %>% 
   filter(!is.na(IDE)) %>% 
  group_by(IDE) %>% 
  mutate(ide_n = n()) %>% 
  ungroup()
 

ide_type <- str_split(df_ide$IDE, ";") %>% 
  unlist()  %>% 
  unique() 

# print("IDEs Included in the survey are")
kable(matrix(ide_type, ncol = 3)) %>% kable_styling()

num_ide <- length(ide_type)
ide_mat <- matrix(ncol = num_ide, nrow = dim(df_ide)[1])
ide_all <- df_ide$IDE

for (i in 1:length(ide_all)) {
  for (j in 1:length(ide_type)) {    
    ide_mat[i, j] = sum(unlist(str_split(ide_all[i], ";")) %in% ide_type[j])
    }  
}

ide_mat2 <- ide_mat %>% as_tibble()
colnames(ide_mat2) <- ide_type

The situation for IDE is similar to languages, it is highly possible that a developer used multiple IDEs at the same time.

ide_mat3 <- cbind(df_ide$Country, df_ide$DevType2, 
                  df_ide$ConvertedSalary, ide_mat2) %>% 
  as_tibble() %>% 
  rename(Country = `df_ide$Country`,
         DevType = `df_ide$DevType2`,
         ConvertedSalary = `df_ide$ConvertedSalary`) 

ide2 <- ide_mat3 %>% gather(Komodo:TextMate, key = "IDE", value = "Value") %>% 
  select(IDE, Value)

ide3 <- ide2 %>% 
  filter(Value == 1) %>% 
  group_by(IDE) %>% 
  mutate(n = n()) %>% 
  ungroup() %>% 
  select(-2) %>% 
  distinct()

ide3 %>% 
  mutate(count = n / sum(n)) %>% 
  ggplot(aes(x = reorder(IDE, count), y = count)) +
  geom_bar(stat = "identity", fill = colors[5]) +
  geom_label(aes(x = IDE, y = 8e-3, 
               label = str_c("(", round(count*100, 2), "%)")),
           fontface = "bold", hjust = 0.5, vjust = 0.5, 
           size = 4, color = "black", fill = colors[3]) +
  scale_y_continuous(labels = percent_format()) +
  coord_flip() + 
  labs(x = "", y = "", 
       title = "IDE Usage") + 
  theme(axis.text = element_text(size = rel(1.15)),
        plot.title = element_text(size = rel(1.15), face = "bold"))

Visual Studio Code* is the king of IDE, with Visual Studio** being the runner up. My favorite is Emacs, which is just as popular as Rstudio, with which I prepared this RMarkdown document.

7.1 Is AI really dangerous?

Let first take a look at the possible answers:

df_ai <- survey %>% 
  select(Country, Employment, DevType2, 
         AIDangerous, AIInteresting, AIResponsible, AIFuture) %>% 
  as_tibble()

df_ai %>% 
  filter(!is.na(AIDangerous)) %>% 
  select(AIDangerous) %>%  
  distinct() %>% 
  kable() %>% 
  kable_styling()

According to the answers to this item, no negative feedback is given.

ai_danger <- df_ai %>% 
  filter(!is.na(AIDangerous)) %>% 
  group_by(AIDangerous) %>% 
  mutate(danger_n = n()) %>% 
  ungroup()

ai_danger %>% 
  select(AIDangerous, danger_n) %>% 
  distinct() %>% 
  mutate(count = danger_n / sum(danger_n)) %>% 
  # print()

  ggplot(aes(x = "", y = count, fill = factor(AIDangerous))) +
  geom_bar(width = 1, stat = "identity") +
  coord_polar(theta = "y", start = 0) +
   scale_fill_manual(values = colors[c(2,1,5,6)])  +
    geom_text(aes(y = c(12, 30, 60, 85)/100,
                  label = percent(count)), size = 4, color = "white") +
  blank_theme +
  theme(axis.text.x = element_blank()) +
  labs(fill = "AI Dangerous?")

7.2 Is AI interesting?

df_ai %>% filter(!is.na(AIInteresting)) %>% 
  group_by(AIInteresting) %>% 
  mutate(count = n()) %>% 
  ungroup() %>% 
  select(count, AIInteresting) %>% 
  distinct() %>% 
  mutate(count = count / sum(count)) %>% 
   # print()

  ggplot(aes(x = "", y = count, fill = factor(AIInteresting))) +
  geom_bar(width = 1, stat = "identity") +
  coord_polar(theta = "y", start = 0) +
   scale_fill_manual(values = colors[c(2,1,5,6)])  +
    geom_text(aes(y = c(86, 20, 65, 46)/100,
                  label = percent(count)), size = 4, color = "white") +
  blank_theme +
  theme(axis.text.x = element_blank()) +
  labs(fill = "AI Interesting?")

7.3 Is AI reposible?

df_ai %>% 
  filter(!is.na(AIResponsible)) %>% 
  group_by(AIResponsible) %>% 
  mutate(count = n() ) %>% 
  ungroup() %>% 
  select(count, AIResponsible) %>% 
  distinct() %>% 
  mutate(count = count / sum(count)) %>% 
  # print()

  ggplot(aes(x = "", y = count, fill = factor(AIResponsible))) +
  geom_bar(width = 1, stat = "identity") +
  coord_polar(theta = "y", start = 0) +
   scale_fill_manual(values = colors[c(2,1,5,6)])  +
    geom_text(aes(y = c(25, 85,  55, 68)/100,
                  label = percent(count)), size = 4, color = "white") +
  blank_theme +
  theme(axis.text.x = element_blank()) +
  labs(fill = "AI Responsible?")

7.4 Will AI have a bright future?

  survey %>% 
  select(AIFuture) %>% 
  filter(!is.na(AIFuture)) %>% 
  group_by(AIFuture) %>% 
  mutate(count = n()) %>% 
  ungroup() %>% 
  distinct() %>% 
  mutate(count = count / sum(count)) %>% 
  
  ggplot(aes(x = "", y = count, fill = factor(AIFuture))) +
  geom_bar(width = 1, stat = "identity") +
  coord_polar(theta = "y", start = 0) +
   scale_fill_manual(values = colors[c(2,1,6)])  +
    geom_text(aes(y = c(50, 95, 12)/100,
                  label = percent(count)), size = 4, color = "white") +
  blank_theme +
  theme(axis.text.x = element_blank()) +
  labs(fill = "AI Future")

Most people are optimistic about, may be the winter is not so close.

8.1 Database Worked with

database <- survey %>% 
  select(Country, Employment, DevType2, ConvertedSalary, 
         DatabaseDesireNextYear, DatabaseWorkedWith) %>% 
  filter(!is.na(DatabaseWorkedWith)) %>% 
  as_tibble() 

db_type <- str_split(database$DatabaseWorkedWith, ";") %>% 
  unlist() %>% 
  unique()

kable(matrix(db_type, ncol=2)) %>% kable_styling()

num_db <- length(db_type)
db_mat <- matrix(ncol = num_db, nrow = dim(database)[1])
db_all <- database$DatabaseWorkedWith

for (i in seq_along(db_all)) {
  for (j in seq_along(db_type)) {    
    db_mat[i, j] = if_else(grepl(db_type[j], db_all[i], fixed = T), 1, 0)
    }  
}

db_mat2 <- db_mat %>% as.data.frame()
colnames(db_mat2) <- db_type

db_mat3 <- cbind(database$Country, database$DevType2, database$ConvertedSalary, db_mat2) %>% 
  as_tibble() %>% 
  rename(Country = `database$Country`,
         DevType = `database$DevType2`,
         ConvertedSalary = `database$ConvertedSalary`) 

database2 <- db_mat3 %>% gather(Redis:Neo4j, key = "database", value = "Value") %>% 
  select(database, Value) %>% 
  filter(Value == 1) %>% 
  group_by(database) %>% 
  mutate(n = n()) %>% 
  ungroup() %>% 
  select(-2) %>% 
  distinct() 

database2 %>% 
  mutate(count = n / sum(n)) %>% 
  ggplot(aes(x = reorder(database, count), y = count)) +
  geom_bar(stat = "identity", fill = colors[5]) +
  geom_label(aes(x = database, y = 2e-2, 
                label = str_c("(", round(count*100, 2), "%)")),
            fontface = "bold", hjust = 0.5, vjust = 0.5, 
            size = 4, color = "black", fill = colors[3]) +
  scale_y_continuous(labels = percent_format()) +
  coord_flip() +
  labs(x = "", y = "", 
       title = "Database Preference") +
  theme(plot.title = element_text(size = rel(1.25), face = "bold"))

Obviously, MySQL is still the most popular database.

8.2 Database Desire Next Year

db_mat2 <- cbind(database2$Country, database2$DevType2, 
                 database2$ConvertedSalary, db_mat2) %>% 
  as_tibble() %>% 
  rename(Country = `database2$Country`,
         DevType = `database2$DevType2`,
         ConvertedSalary = `database2$ConvertedSalary`) 

database2 <- db_mat2 %>% gather(Redis:Neo4j, key = "database", value = "Value") %>% 
  select(database, Value) %>% 
  filter(Value == 1) %>% 
  group_by(database) %>% 
  mutate(n = n()) %>% 
  ungroup() %>% 
  select(-2) %>% 
  distinct() 

database2 %>% 
  mutate(count = n / sum(n)) %>% 
  ggplot(aes(x = reorder(database, count), y = count)) +
  geom_bar(stat = "identity", fill = colors[5]) +
  geom_label(aes(x = database, y = 0.03, 
                label = str_c("(", round(count*100, 2), "%)")),
            fontface = "bold", hjust = 0.5, vjust = 0.5, 
            size = 4, color = "black", fill = colors[3]) +
  scale_y_continuous(labels = percent_format()) +
  coord_flip() +
  labs(x = "", y = "", 
       title = "Database Desire Next Year") +
  theme(plot.title = element_text(size = rel(1.25), face = "bold"))

Like in the database worked with, MYSQL still is the most popular database. The Redis is the most popular NoSQL database in the list.

9.1 Frameworks Worked With

survey %>% select(FrameworkWorkedWith) %>% 
  filter(!is.na(FrameworkWorkedWith)) %>% 
  mutate(FrameworkWorkedWith = str_split(FrameworkWorkedWith, ";")) %>% 
  unnest(FrameworkWorkedWith) %>% 
  group_by(FrameworkWorkedWith) %>%
  mutate(count = n()) %>%
  ungroup() %>% 
  distinct() %>%
  mutate(count = count / sum(count)) %>% 
  # print()

  ggplot(aes(x = reorder(FrameworkWorkedWith, count), y = count)) +
  geom_bar(stat = "identity", fill = colors[5])  +
  geom_label(aes(x = FrameworkWorkedWith, y = 2e-2, 
               label = str_c("(", round(count*100, 2), "%)")),
           fontface = "bold", hjust = 0.5, vjust = 0.5, 
           size = 4, color = "black", fill = colors[3]) +
  scale_y_continuous(labels = percent_format()) +
  coord_flip() +
  labs(x = "", y = "", title = "Framework Worked with") +
  theme(plot.title = element_text(face = "bold"))

9.2 Frameworks Desire Next Year

survey %>% select(FrameworkDesireNextYear) %>% 
  filter(!is.na(FrameworkDesireNextYear)) %>% 
  mutate(FrameworkDesireNextYear = str_split(FrameworkDesireNextYear, ";")) %>% 
  unnest(FrameworkDesireNextYear) %>% 
  group_by(FrameworkDesireNextYear) %>%
  mutate(count = n()) %>%
  ungroup() %>% 
  distinct() %>%
  mutate(count = count / sum(count)) %>% 
  # print()

  ggplot(aes(x = reorder(FrameworkDesireNextYear, count), y = count)) +
  geom_bar(stat = "identity", fill = colors[5])  +
  geom_label(aes(x = FrameworkDesireNextYear, y = 3e-2, 
               label = str_c("(", round(count*100, 2), "%)")),
           fontface = "bold", hjust = 0.5, vjust = 0.5, 
           size = 4, color = "black", fill = colors[3]) +
  scale_y_continuous(labels = percent_format()) +
  coord_flip() +
  labs(x = "", y = "", title = "Framework Desire Next Year") +
  theme(plot.title = element_text(face = "bold"))

Tensorflow turns the 5th framework to get next year, and spark also gained some more popularity.

10.1 Overall distribution

survey %>% 
  select(ConvertedSalary, JobSatisfaction, DevType2, Age, Gender) %>% 
  as_tibble() %>% 
  filter(!is.na(JobSatisfaction)) %>% 
  group_by(JobSatisfaction) %>%
  mutate(count = n()) %>%
  ungroup() %>% 
  select(count, JobSatisfaction) %>%
  distinct() %>%
  mutate(count = count / sum(count),
         JobSatisfaction = factor(JobSatisfaction)) %>%

  ggplot(aes(x = reorder(JobSatisfaction, count), 
             y = count)) +
  geom_bar(stat = "identity", fill = colors[5])  +
  geom_label(aes(x = JobSatisfaction, y = 5e-2, 
               label = str_c("(", round(count*100, 2), "%)")),
           fontface = "bold", hjust = 0.5, vjust = 0.5, 
           size = 4, color = "black", fill = colors[3]) +
  scale_y_continuous(labels = percent_format()) +
  coord_flip() +
  labs(x = "", y = "", title = "Job Satisfaction") +
  theme(plot.title = element_text(face = "bold"))

10.2 Relationship with Salary?

survey %>% 
  select(Country, ConvertedSalary, JobSatisfaction, DevType2, Age) %>% 
  filter(!is.na(JobSatisfaction), !is.na(ConvertedSalary)) %>% 
  group_by(JobSatisfaction) %>% 
  mutate(median_salary = median(ConvertedSalary),
         min_salary = min(ConvertedSalary),
         max_salary = max(ConvertedSalary)) %>%
  ungroup() %>% 
  select(JobSatisfaction, median_salary, min_salary, max_salary) %>%
  distinct() %>% 
  
  ggplot(aes(x = reorder(JobSatisfaction, median_salary), y = median_salary)) +
  geom_bar(stat = "identity", fill = colors[5])  +
  geom_label(aes(x = JobSatisfaction, y = 1e4, 
               label = str_c("(", median_salary, " USD)")),
           fontface = "bold", hjust = 0.5, vjust = 0.5, 
           size = 4, color = "black", fill = colors[3]) +
  # scale_y_continuous(labels = percent_format()) +
  coord_flip() +
  labs(x = "", y = "", title = "Job Satisfaction versus Salary") +
  theme(plot.title = element_text(face = "bold"))

So, are people with salary located in the middle of the salary range struggling?

11.1 Overall Distribution

career <- survey %>% select(CareerSatisfaction, 
                  ConvertedSalary, 
                  Gender) %>% 
  filter(!is.na(CareerSatisfaction)) %>% 
  mutate(gender2 = split_fn(Gender)) %>% 
  group_by(CareerSatisfaction) %>% 
  mutate(count = n()) %>% 
  ungroup() 

career %>% 
  select(CareerSatisfaction, count) %>% 
  distinct() %>% 
  mutate(count = count / sum(count)) %>% 
  # print()

  ggplot(aes(x = reorder(CareerSatisfaction, count), y = count)) +
  geom_bar(stat = "identity", fill = colors[5])  +
  # scale_fill_manual(values = colors[c(1, 2, 5)]) +
  geom_label(aes(x = CareerSatisfaction, y = 3e-2, 
               label = str_c("(", round(count, 3)*100, "%)")),
           fontface = "bold", hjust = 0.5, vjust = 0.5, 
           size = 4, color = "black", fill = colors[3]) +
   scale_y_continuous(labels = percent_format()) +
  coord_flip() +
  labs(x = "", y = "", title = "Career Satisfaction", fill = "") +
  theme(plot.title = element_text(face = "bold"))

11.2 Correlation with Salary?

career %>% 
  filter(!is.na(ConvertedSalary)) %>% 
  group_by(CareerSatisfaction) %>% 
  mutate(med_sal = median(ConvertedSalary)) %>% 
  ungroup() %>% 
  select(med_sal, CareerSatisfaction) %>% 
  distinct() %>% 
  # print()
  
   ggplot(aes(x = reorder(CareerSatisfaction, med_sal), y = med_sal)) +
  geom_bar(stat = "identity", fill = colors[5])  +
  # scale_fill_manual(values = colors[c(1, 2, 5)]) +
  geom_label(aes(x = CareerSatisfaction, y = 1e4, 
               label = str_c("(", round(med_sal/1000, 0), "K USD)")),
           fontface = "bold", hjust = 0.5, vjust = 0.5, 
           size = 4, color = "black", fill = colors[3]) +
   # scale_y_continuous(labels = percent_format()) +
  coord_flip() +
  labs(x = "", y = "", title = "Career Satisfaction with Median Salary", 
       fill = "") +
  theme(plot.title = element_text(face = "bold"))

Generally, better paid careers have significantly higher satisfaction.

12.1 How long can developers be productive?

survey %>% select(TimeFullyProductive) %>% 
  drop_na() %>% 
  group_by(TimeFullyProductive) %>% 
  mutate(count = n()) %>% 
  ungroup() %>% 
  distinct() %>% 
  mutate(count = count / sum(count)) %>% 
  
  ggplot(aes(x = reorder(TimeFullyProductive, count), y = count)) +
  geom_bar(stat = "identity", fill = colors[5])  +
  geom_label(aes(x = TimeFullyProductive, y = 3e-2, 
               label = str_c("(", round(count, 2)*100, "%)")),
           fontface = "bold", hjust = 0.5, vjust = 0.5, 
           size = 4, color = "black", fill = colors[3]) +
   scale_y_continuous(labels = percent_format()) +
  coord_flip() +
  labs(x = "", y = "", title = "Productive Time") +
  theme(plot.title = element_text(face = "bold"))

12.2 Wake Time

survey %>% select(WakeTime) %>% 
  drop_na() %>% 
  group_by(WakeTime) %>% 
  mutate(count = n()) %>% 
  ungroup() %>% 
  distinct() %>% 
  mutate(count = count / sum(count)) %>% 
  
  ggplot(aes(x = reorder(WakeTime, count), y = count)) +
  geom_bar(stat = "identity", fill = colors[5])  +
  geom_label(aes(x = WakeTime, y = 3e-2, 
               label = str_c("(", round(count, 3)*100, "%)")),
           fontface = "bold", hjust = 0.5, vjust = 0.5, 
           size = 4, color = "black", fill = colors[3]) +
   scale_y_continuous(labels = percent_format()) +
  coord_flip() +
  labs(x = "", y = "", title = "Wake Time") +
  theme(plot.title = element_text(face = "bold"))

12.3 Hours Staring Computer

survey %>% select(HoursComputer) %>% 
  drop_na() %>% 
  group_by(HoursComputer) %>% 
  mutate(count = n()) %>% 
  ungroup() %>% 
  distinct() %>% 
  mutate(count = count / sum(count)) %>% 
  
  ggplot(aes(x = reorder(HoursComputer, count), y = count)) +
  geom_bar(stat = "identity", fill = colors[5])  +
  geom_label(aes(x = HoursComputer, y = 5e-2, 
               label = str_c("(", round(count, 3)*100, "%)")),
           fontface = "bold", hjust = 0.5, vjust = 0.5, 
           size = 4, color = "black", fill = colors[3]) +
   scale_y_continuous(labels = percent_format()) +
  coord_flip() +
  labs(x = "", y = "", title = "Hours Using Computer") +
  theme(plot.title = element_text(face = "bold"))

survey %>% select(HoursComputer, ConvertedSalary, Employment) %>% 
  filter(Employment == employed) %>% 
  drop_na() %>% 
  group_by(HoursComputer) %>% 
  mutate(median_salary = median(ConvertedSalary)) %>% 
  ungroup() %>% 
  select(-c(2)) %>% 
  distinct() %>% 
  # print()

   ggplot(aes(x = reorder(HoursComputer, median_salary), y = median_salary)) +
  geom_bar(stat = "identity", fill = colors[5])  +
  geom_label(aes(x = HoursComputer, y = 2e4, 
               label = str_c("(", median_salary, " USD)")),
           fontface = "bold", hjust = 0.5, vjust = 0.5, 
           size = 4, color = "black", fill = colors[3]) +
   scale_y_continuous(labels = percent_format()) +
  coord_flip() +
  labs(x = "", y = "", title = "Longer Hours Facing Computer Better Paid?") +
  theme(plot.title = element_text(face = "bold"))

12.4 Hours Outside

survey %>% select(HoursOutside, ConvertedSalary, Employment) %>% 
  filter(Employment == employed) %>% 
  drop_na() %>% 
  group_by(HoursOutside) %>% 
  mutate(median_salary = median(ConvertedSalary),
         count = n()) %>% 
  ungroup() %>% 
  select(-c(2)) %>% 
  distinct() %>%
  mutate(count = count / sum(count)) %>% 
  # print()

  ggplot(aes(x = reorder(HoursOutside, count), y = count)) +
  geom_bar(stat = "identity", fill = colors[5])  +
  geom_label(aes(x = HoursOutside, y = 3e-2, 
               label = str_c("(", round(count, 3)*100, "%)")),
           fontface = "bold", hjust = 0.5, vjust = 0.5, 
           size = 4, color = "black", fill = colors[3]) +
   scale_y_continuous(labels = percent_format()) +
  coord_flip() +
  labs(x = "", y = "", title = "Hours Outside") +
  theme(plot.title = element_text(face = "bold"))